Dynamic neuro fuzzy diagnosis of fetal hypoplastic cardiac syndrome using ultrasound images

Congenital heart anomalies (CHA) represent a substantial risk to neonates, with 28% to 48% of cases resulting in life-threatening conditions. Consequently, careful prenatal screening is crucial for effective management. Within the spectrum of 18 CHA types, identifying the irregularities in heart morphology, notably the underdeveloped left heart chamber, poses a significant challenge. Hypoplastic Left Heart Syndrome (HLHS), an infrequent yet critical CHA demands diagnosis between the 17th and 21st week of growth. Despite the efficacy of ultrasound imaging, the diagnosis remains intricate due to speckle noise and the complex nature of heart chamber appearances. Selecting an accurate pre-processing algorithm is crucial, and the Fuzzy-based Maximum Likelihood Estimation Technique (FMLET) stands as a pivotal choice. Among the vital parameters for manual diagnosis from ultrasound images, the Right Ventricular Left Ventricular Ratio (RVLVR) and the Cardiac Thoracic Ratio (CTR) play a prominent role. Employing morphological operations such as opening, closing, thinning, and thickening facilitates the extraction of diagnostically crucial features embedded within the images. The development of a Computer-Aided Decision Support (CADS) system, integrating an Adaptive Neuro Fuzzy Classifier (ANFC) proves to be instrumental. ANFC stands out as a better classifier and demonstrates self-learning capabilities similar to that of experts, resulting in a higher diagnostic accuracy rate. The presented Computer-Aided Diagnostic System (CADS) exhibited a notable diagnostic accuracy of 91%, supported by a standardized Area Under the Receiver Operating Characteristic (ROC) curve of 0.92. These results emphasize the system's robustness and effectiveness in diagnosing prenatal CHA, particularly HLHS. Graphical abstract Architecture of the proposed CAD system